An Interval Model Updating Method Based on Meta-Model and Response Surface Reconstruction

Abstract

In this paper, a new interval finite element model updating method is proposed for interval identification of structural parameters based on meta-model and response surface reconstruction. The lower and upper bounds of the uncertain structural parameters are determined by solving the optimization problem which minimizes the difference between the interval of the predicted and measured responses. The response surface models are reconstructed based on the resampling technique for mapping the relationship between a single input and a single output. Then the accurate interval of the responses during the iteration step can be efficiently estimated using the vertex method. Meanwhile, the Gaussian process regression model (GPRM) is constructed as the meta-model to replace the finite element model for calculating the responses of the system to improve computational efficiency. Several numerical and experimental examples are investigated to elucidate the feasibility of the proposed method in the interval identification of structural parameters. Obtained outcomes have demonstrated that the proposed method outperforms many existing approaches in the literature, especially for the nonlinear monotonously non-increasing problem.